Artificial Intelligence in Virtual Screening: Transforming Drug Research and Discovery—A Review
Virtual screening (VS) has become an essential computational tool in drug discovery that helps to identify bioactive compounds by predicting their interactions with biological targets. This approach, which includes both ligand-based and structure-based methods, aims to discriminate between active an...
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
American Association for the Advancement of Science (AAAS)
2025-01-01
|
| Series: | Journal of Bio-X Research |
| Online Access: | https://spj.science.org/doi/10.34133/jbioxresearch.0041 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Virtual screening (VS) has become an essential computational tool in drug discovery that helps to identify bioactive compounds by predicting their interactions with biological targets. This approach, which includes both ligand-based and structure-based methods, aims to discriminate between active and inactive molecules to facilitate the identification of potential therapeutic agents. Despite important advances, challenges remain in accurately predicting ligand–receptor interactions, managing large chemical libraries, and increasing hit identification efficiency. These limitations, coupled with assumptions made by computational tools and data-driven errors, underscore the need for improved techniques to increase prediction precision and reduce false positives. This study employed advanced computational tools for VS, focusing primarily on molecular docking and ligand–protein interaction analysis. AutoDock, known for its Lamarckian genetic algorithm, was used for docking simulations, incorporating pharmacy grids to assess ligand-binding affinities. Additionally, CHARMM software was applied for molecular dynamics simulations to calculate empirical energy functions. AI-driven algorithms such as KarmaDock and DeepDock were utilized for large-scale ligand screening and for improving protein–ligand docking accuracy. Machine learning-based scoring systems and quantitative structure–activity relationship (QSAR) models improved binding affinity predictions. |
|---|---|
| ISSN: | 2577-3585 |